Spaces:
Paused
Paused
Commit ·
4f1e97d
1
Parent(s): 8d99522
updated
Browse files
backend/services/resume_parser.py
CHANGED
|
@@ -77,33 +77,35 @@ def extract_name(text: str, filename: str) -> str:
|
|
| 77 |
# Zephyr Parsing
|
| 78 |
# ===============================
|
| 79 |
def parse_with_zephyr(text: str) -> dict:
|
|
|
|
|
|
|
| 80 |
prompt = f"""
|
| 81 |
-
|
| 82 |
-
Return ONLY a valid JSON object (no extra commentary).
|
| 83 |
|
| 84 |
-
|
| 85 |
-
|
| 86 |
-
|
| 87 |
-
|
| 88 |
-
|
| 89 |
-
-
|
| 90 |
-
-
|
|
|
|
|
|
|
| 91 |
|
| 92 |
Resume:
|
| 93 |
{text}
|
| 94 |
|
| 95 |
-
JSON format:
|
| 96 |
{{
|
| 97 |
-
"name": "
|
| 98 |
-
"
|
| 99 |
-
"
|
| 100 |
-
"
|
| 101 |
-
"education": ["Degree1 - Institution1", "Degree2 - Institution2"],
|
| 102 |
-
"experience": ["Job1 - Company1 (Dates)", "Job2 - Company2 (Dates)"]
|
| 103 |
}}
|
| 104 |
"""
|
|
|
|
| 105 |
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
|
| 106 |
-
outputs = model.generate(**inputs, max_new_tokens=
|
| 107 |
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
| 108 |
|
| 109 |
match = re.search(r"\{.*\}", response, re.S)
|
|
@@ -112,8 +114,8 @@ JSON format:
|
|
| 112 |
return json.loads(match.group())
|
| 113 |
except:
|
| 114 |
pass
|
| 115 |
-
|
| 116 |
-
|
| 117 |
# ===============================
|
| 118 |
# Main Parse Function
|
| 119 |
# ===============================
|
|
|
|
| 77 |
# Zephyr Parsing
|
| 78 |
# ===============================
|
| 79 |
def parse_with_zephyr(text: str) -> dict:
|
| 80 |
+
"""Use Zephyr-7B to extract resume details in JSON format."""
|
| 81 |
+
|
| 82 |
prompt = f"""
|
| 83 |
+
You are an information extraction system.
|
|
|
|
| 84 |
|
| 85 |
+
Extract the following fields from the resume text.
|
| 86 |
+
⚠️ DO NOT return placeholders like "Full Name" or "Skill1".
|
| 87 |
+
Return ONLY actual values from the resume. If a field is missing, leave it as an empty string or empty list.
|
| 88 |
+
|
| 89 |
+
Fields to extract:
|
| 90 |
+
- name
|
| 91 |
+
- skills (list of skills)
|
| 92 |
+
- education (list of degrees + institutions)
|
| 93 |
+
- experience (list of jobs with company, title, dates)
|
| 94 |
|
| 95 |
Resume:
|
| 96 |
{text}
|
| 97 |
|
| 98 |
+
Return ONLY a valid JSON in this format:
|
| 99 |
{{
|
| 100 |
+
"name": "<actual name or empty string>",
|
| 101 |
+
"skills": ["<actual skill>", "<actual skill>"],
|
| 102 |
+
"education": ["<Degree - Institution>", "<Degree - Institution>"],
|
| 103 |
+
"experience": ["<Job - Company (Dates)>", "<Job - Company (Dates)>"]
|
|
|
|
|
|
|
| 104 |
}}
|
| 105 |
"""
|
| 106 |
+
|
| 107 |
inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
|
| 108 |
+
outputs = model.generate(**inputs, max_new_tokens=512, do_sample=False, temperature=0)
|
| 109 |
response = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
| 110 |
|
| 111 |
match = re.search(r"\{.*\}", response, re.S)
|
|
|
|
| 114 |
return json.loads(match.group())
|
| 115 |
except:
|
| 116 |
pass
|
| 117 |
+
|
| 118 |
+
return {"name": "", "skills": [], "education": [], "experience": []}
|
| 119 |
# ===============================
|
| 120 |
# Main Parse Function
|
| 121 |
# ===============================
|